Updated according to ICH GCP R3. 

Coordinating investigator (CI)

In this document, we will use the term Coordinating Investigator (CI) to denote the executive leader of the trial. In a single centre trial, the CI is usually the Principal Investigator. The CI is responsible for:

• providing relevant information to the Trial Statistician (STAT) upon request
• keeping the STAT updated on the study and request advice on arising statistical issues
• reviewing and approving the Statistical Analysis Plan (SAP)
• receive and archive relevant documents in the Trial Master File (TMF)

The STAT is responsible for:

• ensuring that the statistical activities described in this SOP are followed throughout the project and relevant deviations are documented
• ensuring that the statistical activities and analyses are validated as agreed and have the required quality
• creating the SAP
• if applicable, checking that the Data Monitoring Committee (DMC) Charter is adequate and provide code and templates for DMC Statistician (DMC-STAT) and the DMC meeting
• ensuring that the statistical activities reflect the requirements given in the Trial Protocol, the signed Agreement, and as agreed with the CI

The Quality Control Statistician (QC-STAT) is responsible for:

• supporting the STAT in the trial statistical activities
• controlling the quality of the statistical activities
• reviewing and approving the SAP
• be available for statistical enquiries in case the STAT is unavailable

The DMC-STAT is responsible for:

• coordinating input from the STAT and the DMC-STAT to the DMC Charter
• providing the DMC with unblinded analyses and tabulations of interim results based on code and templates provided by the STAT
• participating in DMC meetings and providing statistical interpretations of the interim results
• ensuring the DMC work follows the requirements in the DMC Charter

The Data Manager (DM) is responsible for:

• ensuring that the STAT is involved in the User Acceptance Testing (UAT) as detailed in SOP Data Management . If Data Management is performed by a Clinical Trials Unit (CTU), the CTU's own SOPs will be followed.
• ensuring that the data delivered to the STAT is validated as agreed and have the required quality

Statistical Documentation

The Statistical Documentation for the clinical trial is an electronic archive documenting all statistical procedures such as sample size calculations, randomisation, simulations, SAP (including copies of literature references), statistical code programs and output, and documents for publication (draft manuscripts, submitted manuscripts, review documents, preprints, and published article).

For TMF archiving, a scanned copy of a signature page with all signatures is sufficient and obviates the need to collect and archive the wet-ink signatures. An email confirming the approval may be sufficient.

Emails crucial for documenting the statistical process should be sent for archiving in the TMF. Documents should not contain personal identifiable information from research participants.

Structure and content

Each trial Statistical Documentation is divided into different levels:

Maintenance

The Statistical Documentation should be filed as soon as it is created, unless there are scientific reason to wait (e.g the randomisation list). Documents requiring approval and signature must be signed before filing. Version control should be established with the following specifications:

• Add the document date and version on the first page and in the footer.
• The first draft will have the version number 0.1, with subsequent versions 0.2, 0.3 etc.
• The first final version of the document will be numbered 1.0.
• Subsequent revisions will be numbered 2.0, 3.0, etc.
• Intermittent drafts will be numbered 1.1, 1.2, etc.
• If relevant, a list of changes from previous major version will be added.
• Electronic documents will be stored and named according to:
  •  
  • E.g., GREAT Statistical Analysis Plan v1.0.

Storage and archiving

The STAT will check that all documents are available in accordance with the Statistical Documentation part of the TMF at the end of the study, and the Statistical Analysis Plan is sent to the CI for archiving in the TMF. When the TMF filing is completed, the Statistical Documentation could be kept, if agreed with the CI, as described in the SAP.

Main statistical activity tasks

Protocol development

The STAT will participate in the trial protocol development as agreed. The CI will have the main responsibility for the protocol. Preferably the Trial Protocol should be based on a protocol template provided by the Norwegian Clinical Research Infrastructure Network (NorCRIN). The STAT will critically review the whole protocol and specifically provide input on:

• Trial design
• Choice of outcomes/endpoints
• The DMC (if applicable)
• Data management (together with the DM, if applicable)
• Randomisation and blinding procedures
• Sample size (including replication or new calculations of sample size)
• Statistical methods and data analysis

Randomisation

The STAT will be responsible for generating the randomised allocation list as agreed. The CI will initiate a request for randomisation based on the protocol. The following information will be provided by the CI or designated person:

• Number of treatments
• Total number of participants
• Allocation ratio
• Stratification factors, if applicable
• Degree of blinding
• Template of allocation list

Upon request, the STAT will write the statistical code for the randomised allocation list. The statistical code should have as input a seed for the pseudorandom number generator, and as output the allocation list according to the template. A draft allocation list will be presented to the requesting person for confirmation. Upon confirmation, the final randomised list will be produced as follows:

The STAT will make a request to an independent qualified person to produce the final random allocation list. If the trial requires an independent unblinded DMC and a statistician is appointed to the DMC, the DMC-STAT will produce the random allocation list. The STAT will write the code for the randomisation in the same way as a above and deliver this to the independent person performing the randomisation. The independent person will reset a random seed (e.g., from the random.org website) and run the program. Further handling will be according to request. The STAT is responsible for informing the independent person to keep the final random allocation list unavailable to blinded trial personnel and keep the final program, logs, seed, and output in a secure, access-controlled location until unblinding of the trial. The final program, logs, seed, and output will also be sent to an unblinded Data Manager for uploading into the electronic data capture system (if applicable) and/or to the unblinded person responsible for reporting suspected unexpected serious adverse reactions (SUSARs).

Review of the Case Report Form

The STAT will perform a review of the Case Report Form (CRF) prior to a first finalisation and prior to the release of any further updates affecting the primary endpoint during the study. The review will focus on the primary analysis as specified in the protocol to ensure that all relevant information is collected. The review will at least ensure the collection of:

• all elements involved in the derivation of the primary outcome
• all demographic, baseline characteristics, and other variables included in the primary analysis model
• all elements involved in the definition of the primary study population
• all stratification variables if stratified randomisation is performed

The STAT will ensure:

• together with the monitor and CDM that the system for collection of protocol deviations is adequate
• that the randomisation procedure is correctly entered in the system
• that the STAT, the QC-STAT, and the DMC-STAT are given the right roles and access to the database

Data Monitoring Committee (if applicable)

The DMC-STAT will be a member of the DMC as agreed. The CI will be responsible for appointing clinical members of the DMC and creating the DMC Charter detailing the responsibilities of the DMC. The STAT will contribute to and approve the DMC Charter by email. The DMC-STAT will review and sign the DMC Charter. The DMC Charter must be finalised and signed by the CI and the DMC members prior to inclusion of the first patient.

The STAT will be responsible for providing the DMC-STAT with datasets and code to create the required reports according to the DMC Charter prior to each DMC meeting. The DMC-STAT will acquire the random allocation list and together with the datasets and code provided by the STAT, create the reports, and present them to the rest of the DMC at the DMC meeting. The sharing of information from the reports will be according to the DMC Charter. All information presented to the DMC will be stored in the Statistical Documentation. In blinded studies, the storing will be performed after database lock, keeping the blind for all blinded trial personnel.

The DMC-STAT will ensure that all recommendations by the DMC based on the statistical analyses comply with the protocol and the DMC Charter.

Statistical analysis plan

In ample time (preferably 6 months, but no later than 3 months) before the Data Base (DB) lock or – if applicable – the first efficacy interim analysis, the CI will request the initiation of the trial Statistical Analysis Plan (SAP). This applies both for blinded and unblinded trials. The scope of the SAP will be the main trial publication unless otherwise agreed. The STAT together with the CI and the QC-STAT will meet and reach a decision on principal aspects of the statistical analyses. The STAT is responsible for drafting a first version of the SAP according to Statistical Analysis Plan or the latest TransCelerate template unless otherwise agreed with the CI. The QC-STAT will review the draft prior to presenting the draft to the CI. The STAT will critically review suggestions from the QC-STAT and CI and incorporate changes to the SAP. This process will continue until a final version is reached under consideration of the mock-up report (see section Statistical Programming and Mock-Up Report, including validation). The final version of the SAP will be signed by the STAT, the QC-STAT and the CI before DB lock and un-blinding of treatment allocation.

All trial personnel contributing to the SAP should ensure that the SAP is written without knowledge of the treatment allocation or any results based on this knowledge. This also applies to un-blinded trials.  

All analyses specified in the SAP version signed before un-blinding of the random treatment allocation are denoted “pre-specified analyses”. Any analyses specified after un-blinding of the treatment allocation will be denoted “post-hoc specified analyses”. Due to this, strict versioning control according to section Maintenance

is required for the SAP.  

Statistical Programming and Mock-Up Report, including validation

In parallel with, writing the SAP the STAT will write the statistical code needed to perform the statistical analyses required by the SAP. The code together with a temporary version of the trial database will result in a mock-up report using a fake random allocation to treatment not equal to the actual treatment allocation, regardless of if the trial is open or blinded. This ensures unbiasedness regarding statistical analysis decisions. The code and mock-up report will be reviewed by the QC-STAT and presented to the CI together with the draft SAP for better understanding of the SAP’s implications for the output. The STAT will implement suggestions by the CI and QC-STAT until a final version of the mock-up report is agreed on. The mock-up report does not need to include all output detailed in the SAP but should at least include all results required for the main text of the planned manuscript and reports as specified in the DMC Charter. The report should also include al information necessary for exact reproduction of results, such as a listing of the software version used, the name and version of all relevant supplemental packages or libraries, as well as any seeds for pseudorandom number generators.

The code should be commented if the purpose is not explicitly clear from the code. There should be no hard coding to change the value of observations. In the unlikely instances this is needed, the justification should be clearly stated in an in-line comment. A version control system such as git should be used. All the code should be made publicly available as open source for documentation and transparency. Care should be taken to only publish code and not individual participant data.  

Validation of statistical software packages

The statistical software used is at the discretion of the STAT and will be described in the SAP.

When using commercial off-the-shelf software such as STATA the software needs to be updated to the latest stable version as soon as possible and latest within 6 months of release. Internal functionality should be considered validated against the developed methodology by the vendor, and no specific validation is needed. The use must be in accordance with the user manual.

When using non-commercial open-source software such as R, the base software and all packages need to be updated to the latest stable version as soon as possible and latest within 6 months. Supplemental packages or libraries should only be sourced from central repositories that ensure active maintenance and quality control. For instance R -packages may be sourced from the Comprehensive R Archive Network (CRAN). If developing and/or experimental packages are to be used the documentation and source code needs to be reviewed and assessed against the risk before use.  

Data transfers

When data is transferred from data-collection software to analysis software, checks of data integrity should be performed. Immediately after import to the analysis software, a spot-check must be performed: a date, a numeric value, a categorical value and a string value should be checked against the eCRF.

Data validation

The STAT might be involved in the process of defining the required level of data validation, i.e. various quality checks of the collected data, central monitoring etc., see SOP Data Management 

Blind Review

Prior to breaking the blind, a review of the data and the resulting mock-up report should be done to identify protocol violations, missing data and other issues to be resolved prior to breaking the blind. If the blind review suggests changes to the principal features stated in the protocol, these should be documented in a protocol amendment. Otherwise, it will suffice to update the statistical analysis plan with the considerations suggested from the blind review.

Locking the database

The STAT will be involved in the process leading up to the locking of the database (DB), and when relevant, the assignment of trial participant to analysis populations, see SOP Data Management.

Reporting

Following DB lock and opening of the treatment allocation, the fake treatment allocation will be replaced by the actual treatment allocation in the code to create the final report. The replacement procedure will be reviewed by the QC-STAT. The resulting report will be presented to the CI to be used to write up the manuscript. The report will be version controlled. It is the responsibility of the CI to transcribe the content in the report to the manuscript unless otherwise agreed. If the STAT is deemed eligible for co-authorship according to the Vancouver rules, the STAT has the following obligations in the preparation process of the publication manuscript:

• Provide the text of the Statistical Analysis part
• Quality control of the transcribed content from the report to the manuscript
• Contribute to other parts of the manuscript as appropriate, and critically review and approve the final version for submission
• The CI is responsible to send out the final manuscript for approval by all co-authors
• Contribute to the review process by answering statistical queries and providing requested post-hoc analyses
• Update the SAP with post-hoc analyses

Change control

The STAT is responsible for maintaining the SAP and ensure proper change control during the study.

Statistical programming guidance and training

A statistical guidance document for standard statistical analyses is available at icostatistics.github.io/rct-cookbook. This document should be updated regularly when new methodology is being used. Regular meetings between the NorCRIN members of the statistical group are held where statistical, methodological and programming issues should be raised and discussed. The code available on github should be used for reference when setting up new trials and should be referenced whenever possible. Newly hired statisticians should be trained in statistical programming by an experienced statistician. The training must be documented.

No deviation to this procedure is expected.

• Statistical Analysis Plan - Template
• Protocol - Template
• Data Monitoring Committee Charter - Template
• Data Quality Plan (contact intven@ous-hf.no)